Is this keyword too competitive for your site? A ten-minute check.
Difficulty scores are calculated for nobody in particular. The only question that matters is whether your site can beat the weakest page currently ranking — and you can answer that by looking at the results, in less time than the tool takes to load.
The short version
- Difficulty is relative to you. A score of 30 is easy for an established site and impossible for a three-month-old one. The number is not wrong, it is just not about you.
- Look at position 10, not position 1. You do not have to beat the best result. You have to beat the worst one on the page.
- Forums and Reddit ranking is the strongest green flag there is. It means Google could not find a good page.
- Check for an AI Overview before anything else. If one answers the question completely, the ranking may not be worth having.
- If everything on page one is a household name, walk away. Not forever — just not this month.
Every keyword tool gives you a difficulty score, and every one of them is calculated for a hypothetical average site. Yours is not average — it is either much weaker than that or, eventually, much stronger — so the number tells you where the keyword sits on a general scale and nothing about whether you should write the article.
The check that works takes about ten minutes and requires nothing but the search results.
The six checks, in order
Search the keyword in a private window, ideally in the country you are targeting. Then work down this list. Stop as soon as you hit a decisive red flag.
1. Is there an AI Overview, and does it finish the job?
New, and it belongs first because it can make an otherwise winnable keyword not worth winning. If a summary box answers the question completely, click-through to everything below it falls sharply — reported at around a 58% average decline for top-ranking pages when an AI Overview appears.
The distinction to make is whether the box ends the search. “What is the ideal espresso brew temperature” is finished by three sentences. “Best espresso machine under $500” is not — the reader still wants to be convinced, wants pictures, and wants someone’s actual opinion. Write the second kind.
2. Who occupies the top three?
Not to beat them — to read the room. If the top three are Wirecutter, Consumer Reports and a major magazine, the keyword is a brand-authority contest.
3. Is anything on page one weak?
This is the real test, and it is the one people skip. Scroll to positions 7–10 and look hard at them.
- A forum thread or Reddit post ranking is the single best signal available. It means Google could not find a good article and settled for a conversation. You can write a better page than a forum thread.
- Content last updated three or more years ago, in a category where products change.
- Thin pages — a few hundred words, no images, no author, no method.
- Pages that do not quite match the query — a general “best espresso machines” ranking for “under $500” means nothing specific exists.
- Small sites you have never heard of. If one is there, the door is open.
4. Does the page-one content match what you would write?
If the query returns shopping listings, videos and manufacturer pages, Google has decided this is a transactional query and an article is not what it wants. You can write the best roundup in the world and it will not displace a product grid.
5. How specific is the query, really?
Count the qualifiers. “Espresso machine” has none. “Best espresso machine under $500 for a small kitchen” has three, and each one strips away competitors who wrote something more general.
6. Can you honestly write the best page on this?
Not “a page” — the best one. If you own one of the products, have used them, or can bring a perspective the current results lack, you have something to compete with. If your article would be a rewrite of results one through five, it will rank roughly where a rewrite deserves to. This is where first-hand experience stops being an abstraction and becomes the deciding factor.
A rubric you can actually apply
Score the keyword out of six. One point per green flag.
| Point | Green flag | Red flag version |
|---|---|---|
| 1 | No AI Overview, or one that does not finish the job | A complete answer in the box |
| 1 | A forum, Reddit or Q&A page ranks on page one | All established publishers |
| 1 | At least one result is stale, thin or off-intent | Everything is recent and thorough |
| 1 | At least one small or unknown site ranks | Only household names |
| 1 | The results are articles, matching what you would write | Product grids, videos, retailer pages |
| 1 | You can bring something the current results lack | Your article would be a summary of theirs |
5–6 · Write it now
Genuinely winnable. Put it at the top of the queue before someone else notices.
3–4 · Write a longer variant
Add a qualifier — a price band, a use case, a room, a skill level — and re-run the check on that.
1–2 · Not yet
Not never. Keep it on a list and re-check in six months, once the site has more behind it.
0 · Never
Usually a query where the results are not articles at all. No amount of authority changes that.
Where the winnable ones come from
The check tells you whether a keyword is viable. Finding candidates worth checking is a different job, and the good sources are unglamorous.
- Your own Search Console. Queries where you rank 11–25 are keywords Google already thinks you are nearly relevant for. Improving an existing page is far cheaper than a new one.
- Reddit and forum threads in your niche. People asking questions in their own words, and the fact that a forum was the best available answer is itself the green flag.
- Amazon’s own search suggestions and the questions on product listings. Purchase-adjacent by definition.
- “People also ask” boxes — each one is a query someone typed, often with far weaker competition than the query that produced it.
- Your own single reviews. Model-specific terms have low volume and low competition, and they are the natural first cluster for a new site — see site structure.
The volume trap. A keyword with 100 searches a month that you can rank first for is worth more than one with 10,000 that you will never reach. Ten of the first kind is a thousand visitors a month of highly specific, purchase-ready traffic — and each one is achievable this quarter rather than theoretically achievable in year three.
Two honest caveats
This is a judgement aid, not a prediction. Search results move, competitors publish, and updates rearrange things. The check tells you whether a keyword is worth an attempt today; it cannot tell you the attempt will succeed.
Difficulty scores are not useless — they are efficient for filtering a list of five hundred keywords down to fifty. Use the tool to shortlist and the manual check to decide. What does not work is treating the score as the decision, because it was never calculated with your site in mind.
Part of The Amazon Affiliate Site Playbook — the whole job in six stages, from choosing a niche to keeping a mature site earning.
Find the gaps before you write
Ama Affiliate Pro finds keyword gaps in your niche, checks whether you have already written something too similar, and reads your own Search Console data to surface the queries where you are already close.
Start a 14-day trialCommon questions
How do I know if a keyword is too competitive?
Search it and look at positions 7–10 rather than the top three. If everything on page one is an established publisher with recent, thorough content, it is too competitive for now. If a forum thread, a stale page or a site you have never heard of is ranking, the door is open.
Are keyword difficulty scores accurate?
They are accurate for a hypothetical average site, which is not yours. Use them to shortlist a long list quickly, then judge the shortlist by looking at the actual results. The score answers “how hard is this generally”; you need “can my site beat position 10”.
Why is a Reddit result a good sign?
Because it means Google could not find a good article and settled for a conversation. A forum thread is disorganised, unillustrated and often out of date. A properly structured page with real detail can beat one, which is not true of a well-resourced publisher’s guide.
Should I avoid keywords with an AI Overview?
Avoid ones where the overview finishes the search — simple factual questions. Comparison and recommendation queries survive better, because the reader still wants to be convinced, wants images and wants a human opinion. Check whether the box ends the job before deciding.
How long should I wait before judging whether a keyword worked?
Three to six months for a new site, less for an established one. Pages routinely appear on page three or four first and climb as they gather internal links and engagement. Judging after two weeks tells you nothing.
Is low search volume worth targeting?
Yes, and it is how most small sites get started. A hundred searches a month you can win beats ten thousand you cannot. Ten such keywords is a thousand monthly visitors of highly specific, purchase-ready traffic — and each one is achievable now rather than in year three.
What if the results are shopping listings and videos?
Then Google has judged the query transactional or visual, and an article is not what it wants to show. Write about it if it fits your cluster, but do not expect the ranking. This is the one red flag that authority does not fix.
Where do I find keywords worth checking?
Your own Search Console positions 11–25 first — Google already thinks you are nearly relevant. Then Reddit and forum questions in your niche, “people also ask” boxes, Amazon’s search suggestions, and model-specific review terms, which are consistently the least competitive entry point for a new site.